Control Variables

In addition to these explanatory variables, a set of central variables are included in the baseline statistical model. This is necessary to avoid omitted variable bias. The problem of omitted variable bias emerges when a variable that is correlated with both the dependent and the independent variable is excluded from the analysis. To avoid the problem of omitted variable bias, this paper employs a set of control variables found to be important predictors of severity of conflict and variables often highlighted by similar studies to the study of this paper. Important controls thus include the natural logarithms of population, GDP per capita, incomes generated by natural resources, ethnic diversity, and regime type.

Natural resources are one of the most highlighted aspects of conflict literature. Some main studies, such as studies provided by Collier & Bannon (2003), Fearon & Laitin (2003) and Ross (2003 & 2004) have already created a consolidated knowledge of natural resources as increasing the risk of civil conflict onset. The wealth of natural resources has been argued to materialize its effect through providing finance for rebellion and through influencing the quality and efficiency of a state`s institutions. The availability of natural resources strengthens the incentives of participating the rebellion as such resources present enormous revenues and

61 rewards for rebel groups once the control of these resources is claimed. From a rational point of view, potential revenues and rewards that can be acquired by natural resources exceed the costs associated with participating the rebellion. The effects of natural resources on

motivating people to organize violence against a state contributes also to overcome challenges related to collective action. The private gain to rebels helps them overcome collective action problems (De Soysa & Neumayer 2007, 203).

Another main implication of natural resources is argued to be related to effects of natural resources on state institutions. Natural resources, particularly oil is thought to have corrosive effects on state institutions, allowing patterns of patronage and weak political control (Soysa

& Neumayer 2007, 2004). The term `Resource Curse` is often utilized to label the

implications of natural resources. The term is used to describe the failure of resource-rich countries to benefit from their natural wealth (Humphreys, Sachs & Stiglitz, 2007). The term portrays how countries with substantial natural resources failed to use the wealth of natural resources to improve their economies, and how these countries had lower economic

improvement than countries with scarce natural resources (Moses & Betnes 2017, 5). The lack of state capacity in resource dependent countries increases the risk of conflict by affecting the ability of a state through the corrosive effects of resource wealth on state institutions. To summarize, resource wealth is argued to influence conflicts by creating private material interests for insurgent groups, motivating people to participate insurgent groups because of the rewards that can be provided by controlling the natural resources, financing the insurgent groups to engage in war and finally having erosive effects on state institutions and allowing weak political control. This is consistent with the findings of Weinstein (2007) who argues that access to natural resources creates material benefits and opportunistic rebellions, which are disrupted by indiscipline and tent to commit high levels of indiscriminate violence. Thus, resource wealth and access to natural resources may influence the variation of level of violence in civil wars. To control for the effects of natural resources, this paper incorporates oil rents as a percentage of GDP provided by World Bank data.

Another important control variable is GDP per capita. There are prominently two

comprehensions of employing GDP per capita. Fearon & Laitin (2003) apply GDP per capita as a proxy for state capacity, as low levels of GDP imply incapable and inefficient states, whilst high levels of GDP imply capable and efficient states. In contrast to Fearon & Laitin, Collier and Hoeffler define GDP per capita as a proxy for foregone income. In situations where the foregone income is low, which is measured by GDP per capita in which low levels

62 of GDP per capita purport low foregone income, people may be more likely to participate in rebellions as the opportunities for generating income are very restricted. Collier & Hoeffler (2003) illustrate this by referring to an example of Russian Civil War; Reds and Whites, both rebel armies, had four million desertions. The desertion rate was ten times higher in summer than in winter: the recruits being peasants, the income foregone were much higher at harvest time. Like Feraon & Laitin (2003), Sambanis (2004) also implements GDP per capita as a proxy for state strength. Fearon & Laitin whom emphasises GDP per capita as a proxy for (a) a state`s overall administrative, military, financial and police capabilities, (b) marking

developed countries with better infrastructure and a strong persistence of central

administration in peripheral rural areas and (c) the potential of recruitment as the recruitment of people is easier when economic alternative are worse. Thus, a capable state measured by its economic welfare is associated with responding efficiently to insurgency, lowering

participation by increasing opportunity costs to rebellion and generating institutional strength.

Whether there is the mechanism of foregone income, the institutional strength, or a capable state, we may expect economic wealth to reduce the intensity of a conflict. To control for the economic wealth, this paper incorporates GDP per capita provided by World Bank data.

Ethnic diversity is another variable that is often included in the studies of conflict literature. In literature the main interest of incorporating the ethnic diversity has been to examine if

ethnically diversified countries were more prone to conflicts. Most of these studies do not find any clear and strong effects of ethnic diversity on either the risk or durability of a conflict (Fearon & Laitin 1996; Hegre & Sambanis 2006; Reynal-Querol 2002; Sambanis 2001). The question of whether ethnicity is indeed a factor that leads to make a conflict more severe has not obtained a complete scrutiny. Given the existence of an ongoing conflict, may ethnic diversity increase the severity of a conflict? The role of ethnicity in conflicts and how ethnicity should be expected to influence severity of a conflict and other aspects of conflicts generally, may be related to the level of aggregation. When interpreted at country level, ethnic diversity may examine if the conflicts are severe in ethnically diversified countries. There is little guidance in the literature of severity of conflict on how ethnicity and ethnic diversity may influence the severity of conflict. Of the few studies that attempt to elaborate

determinants of severity of a conflict, Lucina (2006) finds no indication that cultural diversity makes a civil war more severe, Lu & Thies (2011) finds no support for three indicators of cultural grievance, that is, the share of the second largest ethnic group, ethnic-linguistic, and religious fractionalization, Balcells & Kalyvas (2014) finds ethnic fractionalization reduces

63 battlefield severity and Lujala(2009) finds also that ethnically diverse countries seem to experience conflict with fewer battle-related deaths. These studies provide some mixed results on how ethnic diversity may affect severity of a conflict. We also infer from these studies that there is lack of profound theory and convincing arguments why and how ethnic diversity should influence severity of a civil war. In addition, the studies employed ethnic diversity at country level. In contrast to these studies, Eck (2009) attempts to a give a more

comprehensive examination of ethnic diversity. Eck emphasizes ethnical dimension on both country level and rebel group level and provides a more convincing argument of how ethnically mobilized rebel groups may be more likely to intensify a conflict than groups that do not mobilize along ethnic lines. Thus, we may argue that the choice of employing ethnic diversity either at country level or rebel group level may affect the results of ethnic diversity.

This paper employs the conventional use of ethnolinguistic fractionalization provided by Fearon & Laitin. The variable measures the probability that two randomly selected people from a given country will belong to different groups. The variable ranges from 0 (perfectly homogenous) to 1 (highly fragmented). Note that this variable is at country level and may not be best suited to account for the effect of ethnicity and ethnic diversity. A similar strategy to Eck who conceptualizes ethnicity at rebel group level and in terms of the role of ethnicity in mobilizing insurgent groups might provide more significant and accurate results. Ethnical diversity also has implications for the context of elections. In ethnically diverse countries the politics, voting and issues of interest during elections may primarily highlight ethnic rivalries (Long & Gibson 2015, 830). Ethnicity, which is an important source of social differentiation such as race, plays a considerable role in shaping people`s political attitudes and voting behaviours (Just 2017). Elections that fail to reflect the ethnical diversity of a society may not be perceived as legitimate and representative, which thus influences the quality of an election.

The regime type will also be included as a control variable in this paper. The regime type of a country may affect both elections, characteristics of elections and severity of a conflict.

Regime type of a country is related to if a given country holds an election or not and related to the attributes of elections in a country. The elections in democracies are free, fair and

multiparty elections, whilst in non-democratic settings there may be multiparty elections but the probability of government change may be very small as there is not a real competition and the incumbents may be favoured against others. There is also a considerable variation of regime types. Distinguishing regime types across democracies versus non-democracies may not be feasible for controlling the variety of regime types. This paper employs a four type of

64 regime variable; democracy, military, monarchy and multiparty. The conceptualizations of these regime types are provided by Hadenius, Teorell & Wahman (2013) who attempt to develop, and detail two other widely used classifications, those introduced by Geddes et al (2012) and Cheibub et al (2010). Hadenius, Teorell & Wahman (2013) conceptualize

monarchies; «as those in which a person of royal descent has inherited the position of head of state in accordance with accepted practice and/or the constitution». Cases in which the monarch possesses limited political powers and in not the effective head of government are not classified as monarchies. These are ceremonial or constitutional monarchies. Military regimes are defined as: «states in which military officers are major or predominant actors by virtue of their actual or threatened use of force». In these regimes, the armed forces control politics directly or indirectly by directing civilian leader behind the scenes. To acquire a definition of democracy, Hadenius et al refer Freedom House and Polity scale, which are two of the most used indicators of democracy. Their main incentive is to find a threshold that could provide them with a tool to define a country as a democracy. To find this threshold, they combine Freedom House and Polity scale. Thus, they classify all countries with a democracy score of 7 or above as democratic. In the upcoming paragraph, an elaboration of how Freedom House and Polity IV conceptualizes democracy is presented. Both indicators are used by Hadenius et al to provide a scale for democracy.

The last category of regime type that this paper employs, is multiparty regimes. Multiparty regimes will be controlling a regime setting that is becoming quiet dominant. The category corresponds to the regimes labelled as «electoral authoritarian» by Schedler (2006) or

«competitive authoritarian regimes» by Levitsky & Way (2010). The common aspect for these categorizations is that a certain level of competition is allowed, and some opposition candidates can participate in national elections. Elections are held in these settings, but the chance of the opposition to win the elections is almost non-existent.

The last control variable of this paper is the size of population, which is a typical variable in conflict intensity-related research. The population is the source of both insurgents and government soldiers. Countries with a rich source of combatants due to large populations result in rebel and government commanders who are less concerned about casualties, since combatants are replaceable (Lu & Thies 2011, 223). Following this logic, we may assume larger populations to intensify a civil war and make it more severe. Previous empirical works provided by Lucina (2006), Lujala (2009) and Lu & Thies (2011) support this claim. To

65 control for the effect of population size, this paper employs a logged variable of population size obtained by World Bank.